AI Skill Dojo
Friends & Family · By Personal Invitation
Free · You're on the list · Here's what you signed up for

My gift to you: let's learn AI together.

You already said yes in the group chat — so this isn't an invitation, it's the briefing. I coach professionals in AI skills, and I'm running that same curriculum for my own people as a free live workshop: small sessions on Zoom — one or two of you at a time — two 2-hour sessions, the same material my paying students get. Not a sample. Not a trial. The real thing — because you're my people, and because, honestly, you'll be teaching me something too.

Free
A gift, not a funnel
2
Sessions · 2 hours each
Zoom
1–2 people per slot · you pick the times
W→P
White→Green + Purple demo
$20
The plan we'll max out
0
Coding or tech needed
The honest deal, up front

You get: real, hands-on AI skills — the same method and materials my paying mentees get, at whatever pace fits your life, on whatever you actually care about: work, a side project, travel planning, the family photo archive, anything.

I get: a teacher's most valuable thing — honest students. Every session with you teaches me how AI actually lands with real people outside the tech bubble: what confuses, what excites, what feels creepy, what your workplace allows, what you'd genuinely pay for. That knowledge directly shapes what I teach next and to whom. You're not a guinea pig — you're a co-designer. And I'll tell you exactly what I'm learning from you as we go.

01What Each of Us Walks Away With

You walk away with

Real skills, not tech-bro hype

  • Use AI properly — get consistently useful answers instead of generic fluff, on your own tasks
  • Privacy configured — the settings most people never find, set up on your own accounts
  • Max out the $20 plan — unlock the agent surfaces, projects and connectors already inside the subscription you're probably underusing
  • Stop overpaying — know which AI tool (free or paid) fits which job, before spending another dollar
  • Cross-check instead of trusting — put the same question to a second AI from a different company, ask for sources and actually open them, and learn which questions are risky in the first place: fresh news, exact numbers, and anything sitting behind a login the AI was never allowed to read
  • Something you built — every stretch ends with a real thing you made and keep
You'd be teaching me

What the real world needs

  • How people of your generation actually feel about AI — the hopes and the ick, unfiltered
  • What your workplace expects: what skills employers actually want, and what training (if any) they provide
  • Which AI tools your company's IT actually allows — and which are banned
  • What AI is genuinely worth to you — would you pay $0, $20, or $200 a month, and for what
  • Where my teaching confuses or bores you — the feedback paying students are too polite to give
My privacy promise — the part that matters in a family

02What You'll Learn — The 5D Method

My own design, and the spine of everything I teach: the five things a company does to ship a product — Discussion, Design, Develop, Deployment, Debugging — compressed onto one desktop, with AI doing the heavy lifting and you keeping the judgment. You don't need to code. You don't need to be "technical." You need curiosity and a browser.

D1 · Discussion

Research Team

Get answers that land right the first time

D2 · Design

Product Office

Check the facts, decide what gets built

D3 · Develop

Engineering Team

AI builds from your instructions

D4 · Deployment

Launch & Ops

Test it, then make it real

D5 · Debugging

Failure Analysis

Find out why, when AI gets it wrong

Three skills that carry it — practiced on your real life, not toy examples

Foundation

Meta-Prompting

Let the AI interview you before it answers — and keep the verification habits that stop you trusting it blindly. The difference between generic fluff and output you can actually use.

The multiplier

Context Engineering

Teach the AI who you are, how you work, and what you're really asking — including handing it your own documents so it answers from your material instead of guessing from whatever it half-remembers. This is what turns it into a competent helper for your life and job, and it keeps teaching you long after our sessions end.

The leap

Agent Bootstrapping

Go from chatting with AI to directing it: hand a whole job to an AI agent, let it work, and verify what comes back. This is the part most people have already paid for and never use.

How far we go in two sessions

The full program runs five belts. Our two sessions take you hands-on through the first three — everything most people need day to day — and close with a look over the fence at the fourth.

White Belt
Session 1
Use AI properly, privately, and without overpaying. Meta-prompting, privacy settings, which tool for which job, verification habits — and an artifact you build live in the session.
Yellow Belt
Between sessions
Make AI genuinely yours. Context engineering on something real from your own work or life — so the answers stop being generic and start being for you.
Green Belt
Session 2
Direct the work instead of doing it. Hand a full research job to an AI agent, watch it run, then verify the result — including catching at least one confident wrong answer together.
Purple BeltLive demo
Session 2 · finale
Watch a one-person company run. To close, I'll drive a live multi-agent orchestration across Claude Desktop's three surfaces — Chat, Cowork, and Code, plus Claude Code CLI on a server — several AI agents working one job together while one person supervises. You watch this one rather than drive it — and by then you'll understand exactly what you're watching.

We start wherever you are — total beginner is the most common starting point and honestly my favorite. And if all you want from the whole thing is Session 1's "make AI useful and stop worrying about it," that's a complete, perfect outcome too.

If you're skeptical about AI — especially you

Maybe you think AI is overhyped, creepy, coming for jobs, or all three. I'm not going to talk you out of any of that — some of it is true, and the people most worried about AI are usually the ones paying closest attention.

What I'll show you instead is where it breaks, where it lies, what it should never be trusted with, and how to use it on your terms with your privacy intact. Skeptics make my best students — and your skepticism is, sincerely, part of what I want to learn from.

03Get Everything You're Already Paying For

Here's the thing almost nobody knows: the $20-a-month AI subscription most people already have is a professional-grade toolkit, and the overwhelming majority of subscribers never leave the chat box. They're paying for a workshop and using the doorbell. A large part of these two sessions is simply showing you what's already yours.

The surfaces you've never opened

Beyond the chat window sit agent workspaces that take a whole job and go do it — research, files, spreadsheets, multi-step work. Same plan. No upgrade. Most people have never clicked them.

Projects — your own walled garden

A permanent workspace stocked with your documents, your style, your standing instructions. Ask a question there and it answers from the material you handed it rather than from a hazy memory of the internet — the industry calls this RAG, and it's the most dependable way to keep answers anchored to something real. You stop re-explaining yourself, and it stops guessing.

Connectors & file work

Hook it up to the places your actual work lives, hand it real documents, and let it produce real files back — not just text you copy-paste and reformat by hand.

Model routing & cost sense

Knowing which model to reach for — and when the cheap fast one is genuinely the right answer — is most of the difference between a $20 plan that feels tight and one that feels generous.

Why we start on Claude Pro — and why it doesn't lock you in

We'll all work on Anthropic's Claude Pro for these sessions, for one practical reason: Pro already includes the advanced features — the agent surfaces, the projects, the connectors, the developer tooling. There's no "but that's the higher tier" moment mid-lesson. If you ever need more headroom, you simply top up credits rather than buying your way up a ladder of plans. For a class, that means everyone sees the same screen and nobody gets fenced out.

The skills transfer — including off the paid track entirely. What you learn here — how to frame a request, how to give the AI proper context, how to hand off a whole job and then verify what comes back — works on ChatGPT and Gemini, and it works just as well on the open-source stacks: OpenClaw, Hermes, and open-weight models you can run on your own machine with no subscription at all. The buttons move; the method doesn't. I teach the method precisely so you're never stranded when your employer standardizes on a different vendor, when a budget gets cut, or when something better launches next year. Loyalty to a method, not to a logo.

The cheap seats — worth knowing

Once you know what you're doing, you'll notice most everyday AI work doesn't need the most expensive model available. Matching the model to the job is most of the difference between a plan that feels tight and one that feels generous — and some of the best value sits in newer, far cheaper models most people have never heard of.

Which ones — and the rules for using them safely — I'll post to the group chat separately. They're genuinely useful and genuinely conditional: where your words end up, what your employer allows, and what should never go near them. That belongs in a conversation with rules attached, not a line on a page.

The tools you'll hear me name

No mystery jargon. These are the actual pieces on the table, and where each one sits.

Claude Desktop · Chat
Frontier · included
The part everyone knows — but used properly. Think and research here. Where meta-prompting and context engineering live.
Claude Desktop · Cowork
Frontier · included
The agent workspace. Delegate whole jobs here — it works with your files and reports back. This is the one hiding in plain sight inside the plan you already pay for.
Claude Desktop · Code
Frontier · included
Build things here — and yes, without knowing how to code. You describe and supervise; it types. We'll touch this lightly in the demo.
Claude Code CLI
Frontier · advanced
The same builder running headless on a server, no window at all. Demo only in our sessions — this is where the 1-on-1 road starts.
Gemini Spark
Frontier · Google · $20 tier
Google's always-on agent — a butler for the Google side of your life. It keeps working on Google's servers after you close your laptop, takes jobs by email, and reaches across Gmail, Docs, Drive and Calendar; it'll handle standing chores like surfacing deadlines buried in your inbox. I name it for a reason that matters more than the features: it arrived on the $20 tier too. Two rival companies, same story — you are probably already paying for an agent you've never once used. (US on the $20 plan as of Aug 2026; some of it is still on the higher tier.)
Claude API
(developer platform)
Direct · pay-per-use
The same models, reached directly instead of through an app — and billed only for what you actually use, often fractions of a cent per request, with no monthly plan at all. This is the layer any real AI product is built on. The capstone I'll demo runs on exactly this, which is why it costs pennies to operate.
LLM brokers
(multi-model gateways)
Aggregator · one key, many models
A middle layer worth knowing about: one account and one key that reaches dozens of models from many different companies — frontier, open-source and low-cost alike — where switching between them means changing a single word. Excellent for budgeting, and the easiest way to put the same question to two rival models when you want a second opinion. The trade: your requests now pass through one more party, so the jurisdiction and policy questions above apply to the broker as well as to the model. I'll name the one I use in session.
GitHub
(repository & auto-checks)
Platform · free tier
Where a project actually lives: every version kept, every change traceable, and free robots that re-check your work each time you touch it. It sounds like a developer thing and it isn't — it's the difference between "I think this still works" and being able to prove it does. My capstone lives in one, and you'll see its automated checks run during the demo.
VS Code
IDE · free
The free editor where building actually happens — and where Claude's Code surface plugs in, so you can watch it work on real files instead of trusting a summary. You will not be asked to write code in it. Treat it as the workshop window: you see what's being made, what changed, and you can stop it mid-job. Same editor the professionals use, same price as nothing.
Copilot
Frontier · often employer-supplied
Named here because for several of you this is the AI your workplace already pays for — and possibly the only one IT permits. Two different products share the name: the one woven into Microsoft 365 (Word, Excel, Outlook, Teams) and the one for writing code on GitHub. Everything in these two sessions — how to frame a request, how to supply context, how to verify what comes back — applies to both. Bring me your IT policy and we'll work inside what you're actually allowed to use.
OpenClaw
Open source · demo
An open-source agent runtime. Useful as an X-ray of how an agent is actually built — the anatomy the polished products hide from you.
Hermes
Open source · demo
Another open agent stack, shown side by side with the commercial ones so you can see what you gain and what you give up going open.

Why show the open-source ones at all when the paid tools are easier? Because seeing the same idea in two very different implementations is what turns a tool you memorized into a concept you understand — and because knowing what's free matters when a budget gets cut or an employer says no. The same goes for the pay-per-use and broker layers: once you've seen that the monthly plan, the direct API and the gateway are three doors into the same room, you stop being a customer of one product and start choosing on the merits.

04Who's Teaching — In Case You Only Know Me As Family

Some of you have known me forever and have no idea what I actually do all day. Here it is, briefly.

Mike Cho · Senior Project Engineer & AI Coach · Bay Area, CA · 25+ years
Hardware validationSoftware QANetwork engineering Wireless & telecom R&DSilicon Valley labs

Twenty-five years as an engineer in Silicon Valley — hardware validation, software QA, and network engineering across R&D labs and consulting contracts where "we'll ship it next quarter" was never an acceptable answer. My job, for a quarter century, was finding out why things break.

I've lived through four technology disruptions: the dot-com bust, the telecom collapse, outsourcing, and offshoring. AI is the fifth, and I'm getting through it the same way I got through the others — by learning faster than the disruption moves. That system is what I'm teaching you. The engineers who survived each wave weren't the ones who memorized a platform; they were the ones who had a method for putting new tools to work quickly.

Sessions in English · Mandarin · Cantonese  ·  Zoom or Google Meet  ·  US and Asia-Pacific friendly hours

05How It Works

1

You're already in

You said yes in the group chat. There's nothing to sign and nothing to pay — that part is finished.

2

You book your own slots

Getting everyone onto one weekend turned out to be impossible — some of you can only do weekdays, some only weekends. So there's no single date any more. I'll send a Calendly link to the group; pick the two times that actually suit you.

3

A little prep

Bring a laptop, not a phone — you'll want a real keyboard. Have a free AI account ready if you can; if that sentence worries you, message me and we'll sort it in five minutes.

4

Two sessions, 2 hours each

Hands-on from minute one — you drive your own keyboard while I coach. White through Green Belt, closing with the live demo.

5

You tell me the truth

What worked, what confused you, what your workplace really thinks of AI. That's the gift back — and it's plenty.

Zero obligation — actually zero

Life happens. If you make the first session but not the second, or drift off halfway, nothing changes between us. If you'd rather just have coffee and never talk about AI again, we're square. This is a standing offer, not a commitment — the relationship always outranks the project.

Want to drive it yourself? That's the 1-on-1 road

Purple hands-on — and Brown beyond it — is where the work leaves the chat window entirely: your own server on the internet, run by agents you command, hosting something real that you built. That's how I built my Award Transfer Optimizer, with the same method you'll learn in our two sessions.

It can't be group work, because the environment has to exist first: a Linux cloud VM with remote desktop — a second computer of your own that lives in a data centre instead of on your desk, always on, used from your laptop. Standing it up is realistically two sessions, done together at your pace, and none of it is knowledge you need in advance.

Most people won't need this, and that's genuinely fine — the two sessions stand on their own. It isn't part of the gift and it isn't group work: it's a separate arrangement, made offline, one person at a time. If it's the direction you want to go, we can arrange for 1-on-1 sessions.

You're in. Now just book your two slots.

One big group on a single weekend was never going to work — some of you can only do weekdays, others only weekends. So you pick your own times instead. I'll send a Calendly link to the group chat: choose two 2-hour slots that suit you, and either come on your own or put two of you on the same booking. You're the first batch, so you get first pick of the calendar.

Questions before we start, or something you're quietly hoping we'll cover? Send it to the group, or to me directly. Requests shape the sessions — that's not a courtesy, it's how I find out what to teach.

Free for close friends & family · booking link goes to the group chat · Zoom link comes with your confirmation
Two 2-hour sessions, booked to suit you — weekday or weekend, one or two people per slot — closing with a live multi-agent demo · upper belts by separate 1-on-1 arrangement, offline
English · Mandarin · Cantonese